Real Time SOC Estimation Based On EKF And UKF

“Safety, energy saving and environmental protection” is the eternal theme of automotive technology development. In today's global energy crisis and environmental problems are becoming increasingly serious, new energy vehicles, represented by electric vehicles, have become the future development trend of automobiles. As one of the key technologies of electric vehicles, the safe and efficient use of batteries depends on the accurate estimation of battery status. The state of the battery can be divided into two categories, one can be directly measured, such as voltage, current, temperature, etc. The other category cannot be directly measured, but can only be estimated by certain methods, such as the state of charge SOC, state of health SOH and so on. These state quantities are critical in the process of battery use. There are many methods for estimating SOC, and most of them are based on the basic principle of ampere-time integration. Although the integration method is simple and easy to implement, there are two important problems: (1) the initial value of SOC cannot be estimated; (2) the inaccuracy of current measurement will cause cumulative errors. To solve these problems, the Kalman filtering method can be used. For our battery, we build two models, EKF and UKF, to predict its SOC variation and compare them respectively. In this paper, the third-order Thevenin model was chosen to describe the dynamic behavior of the battery after comparing various equivalent circuit models and considering the accuracy and complexity of the model. The model is used to identify the parameters of the battery through pulse current charging and discharging tests. The results show that the Thevenin model can better describe the dynamic behavior of the battery, and its structure is simple and easy to identify the parameters. Then, the battery SOC is estimated based on the traceless Kalman filter (UKF) algorithm in Matlab environment. The accuracy and convergence speed of the algorithm are verified by artificially creating initial value errors and input noise by using WTP3 conditions to verify the effectiveness of the algorithm in a single working cycle. The simulation results show that the algorithm can gradually converge to the true value of SOC and follow it when the initial value is inaccurate and the input contains noise, and the estimation error does not exceed 4% in the middle and late stages of the working condition test.